What problem does it solve? It prevents scope creep and uncontrolled edits when an AI agent makes small code changes by enforcing a strict brief with authorized paths, pre-verified facts, and a hard stop after verification. ## Core Features & Use Cases - Bounded Execution: Works only inside AUTHORIZED PATHS, trusting KNOWN FACTS without re-exploration, and makes exactly the one CHANGE requested. - Verification Gate: Runs the CHECK commands as written, allows one repair attempt, and stops after two failed verification runs. - Escalation Discipline: Stops and reports instead of improvising when scope grows, a KNOWN FACT is wrong, or the work needs a full plan or work order. - Use Case: A coordinator hands the agent a brief to rename a function in two specified files and run the test suite; the agent makes only that edit, runs the tests, reports the result, and stops. ## Quick Start Hand the agent a quick-executor brief containing GOAL, AUTHORIZED PATHS, KNOWN FACTS, CHANGE, CHECK, and ESCALATE IF sections and ask it to implement the single change.